AI-First Strategy

In an AI-first strategy, AI operates at the core of a company, driving its product and decision-making. In several industries, new challengers are using this kind of strategy to successfully compete against incumbents. One example is in the financial industry, with companies like Citadel, Two Sigma, and Personal Capital maximizing return and reducing risk by creating the best machine intelligence. And in the automotive industry, it’s now easy to envision a day when people will decide to buy a new car based on its driving software and not on its engine, body design, or other buying criteria.

Across all industries, the use of deep learning has the potential to increase production, drive down cost, reduce waste, and improve efficiency, as well as push innovation. And, as can be seen in industry leaders like Google, Apple, and Amazon, machine intelligence changes everything and becomes pervasive when an organization pieces together how to use it.

Though we cannot predict its full impact, it’s clear that AI represents profound change both in the short and long term, and it is a technology that demands strategic focus and action.

Where Deep Learning Excels

Deep learning is not the best approach for every problem, and more basic tools such as stats and machine learning aren’t going away anytime soon. But deep learning is extremely powerful when we have access to plenty of training data, we have many dimensions or features of the data (which would require time-consuming feature extraction in order to conduct machine learning), or we need to process rich media such as images, video, and audio.

In the short term, the techology holds significant promise for dealing with problems like fraud detection, predictive maintenance, recommendation engines, yield optimization, and churn reduction. In these areas, it could produce order-of-magnitude improvements in two ways:

In some cases, deep neural networks will yield better predictability over current models even when using the same dataset

Later in the book, we discuss how deep learning models were able to detect fraud much more predictably than machine learning ones at Danske Bank, even using the same data as prior-generation models.

Deep learning can allow the enterprise to analyze previously intractable datasets

For example, companies could use images and audio files for predictive maintenance, as in mining photos of a piston in an engine to spot cracks and other imperfections before they become more serious or using audio from wheel bearings in a train to listen for anomalies that signal a potential derailment.

Let’s take a look at some use cases for which deep learning could provide a significant advantage over current prediction methods.

Financial Crimes

Financial crimes cost institutions, consumers, and merchants billions of dollars every year. In the past, fraud was more difficult to perpetrate because banking was personal and channels for crime were more constrained. The internet has changed that. Modern banking is almost completely anonymous and occurs through many avenues. This has enabled a new ecosystem of many kinds of fraud, which are growing increasingly sophisticated and aggressive. Both industries and governments face unprecedented threats from a variety of actors, risking physical loss of money, intellectual property theft, and damage to their reputations.

Financial institutions have long been using machine learning, data mining, and statistics to mitigate risk, and these have certainly provided value. Today’s risk landscape, however, demands new tools.

Though deep learning didn’t initiate real-time or cross-channel data analysis, it is better at detecting more accurate patterns across all data streams, in addition to its ability to analyze new types of data. Because of this, AI can empower banks and other institutions with insight that keeps up with the pace of modern fraud.

Manufacturing Performance Optimization

Currently, the manufacturing industry suffers from inefficiencies due to “siloed” data and delayed communication of insights across the supply chain, from the acquisition of raw goods through production and sales. Improving this efficiency represents a huge opportunity for manufacturers.

Iterating on manufacturing processes is nothing new—it’s something that has been done for decades. However, AI can permit iterations and adjustments to systems in minutes instead of months.

The increased predictive power of AI enables companies to proactively understand their needs and intelligently communicate them across their different branches. This can have a huge impact on every part of the business. According to data from General Electric, smart manufacturing systems using AI can increase production capacity up to 20% while lowering material consumption by 4%.

AI provides data to the business in real time, which can help optimize the supply chain, provide greater economies of scale, and better manage factory and demand-side constraints. GE, for example, saw finished goods buffers reduced to 30% or more by using a smart manufacturing system.

There are many other manufacturing use cases for AI, such as intelligent pricing, ensuring regulatory compliance, improving eco-sustainability, and finding new revenue streams. As the technology develops, more use cases are sure to be discovered.

With its abundance of sensor data and systems that richly reward increased efficiency, manufacturing is an industry poised to be revolutionized by machine intelligence.

Recommendation Engines

Whereas companies used to get to know their customers’ buying habits and preferences through face-to-face interactions and relationships built over time within the four walls of a store, they must now infer this same data through online activity. This has made recommendation engines essential for many businesses to compete in the online marketplace. By helping customers discover items or content quickly, recommenders increase satisfaction, expenditures, and lifetime value.

Even though recommendation engines predate deep learning, their effectiveness continues to grow for retailers who have made the switch from legacy engines driven primarily by collaborative filtering to ones based on wide and deep learning. For example, Amazon reported that an impressive 35% of sales are a result of their recommendation engine, and 75% of content watched on Netflix comes from such algorithms.

We can use deep learning algorithms at several points for building a recommender. They allow radical personalization, enabling each person to see items particular to their interests and actions, thereby solving the difficult problem of how to show one out of thousands of customers the right product out of thousands of options.

Deep learning can also find unique connections across items that might not be intuitive, such as showing other baby products when someone is searching for children’s books, or showing a user novel items, creating the feeling of serendipity. When done well, recommenders essentially scale the record store clerk or the friendly sales associate, helping customers find both what they want and what they didn’t know they wanted.

Yield Optimization

A manufacturer must consider many variables when determining how much product to produce. These can include supplier, customer, and team requirements as well as equipment availability and capacity. Unfortunately, factories often perform at less-than-optimal rates due to imperfections and lag-time in communication between devices and management. This includes data streams that are disconnected from one another and variables that often change but require manual processes to take their new values into account.

Manufacturers across industries like aerospace, high-tech, and industrial equipment have been using AI to improve communication across their devices and are seeing gains in yield, creating more efficiency and using more of their production capacity.

Yield optimization presents a huge opportunity because even incremental changes in efficiency can create significant value. For example, leaders at Micron found that each 1% cumulative yield improvement translated to $100 million in annual cost reduction. They were then able to use AI with sensor data to determine the top factors negatively affecting yield, ultimately leading to significant financial and operational improvements.

Predictive Maintenance

According to Statista, predictive maintenance is one of the top-ten most valuable use cases for AI, with the potential to generate $1.3 trillion by 2025. This makes sense, as unplanned downtime is extremely expensive, costing manufacturers billions of dollars every year. In the automotive industry alone, each minute of unplanned downtime costs $22,000.

Predictive maintenance with AI is a key part of factory automation. It helps reduce capital expenditure, extends the lifetime value of equipment, and improves safety. With machine intelligence, manufacturers can more efficiently integrate and analyze across all data sources, improving the performance of maintenance, repair, and overhaul (MRO), as illustrated in Figure 2-1. It also manufacturers them greater insight into each component and part, allowing potential points of failure to be identified and repaired before they become a problem.

Figure 2-1. Business outcomes enabled by Industrial IoT predictive maintenance

Traditionally, manufacturers were predicting downtime by building algorithms fed on data sources like maintenance records, parts inventory, and warranties (small data), as well as big data like sensor streams coming from a jet aircraft engine or an MRI machine. These models functioned fairly well, and these algorithms were able to use that data to make much more accurate predictions than models that did not.

Figure 2-2 illustrates how deep learning improves on that technology by enabling the use of new types of data like audio and video that can be incorporated with traditional data sources to enhance prediction capabilities. In essence, deep learning scales the task of someone taking a look at hoses to check for cracks or listening for strange sounds on the shop floor.

Figure 2-2. Deep learning augments traditional data sources and analytic techniques

We have not yet come close to tapping the full potential of deep learning. We will discover more use cases as the technology continues to develop and is implemented across all industries.

In Chapter 3, we discuss ways that enterprises are currently able to consume AI and build deep learning capabilities.